Friday, 30 November 2007

COTs Look Sunny for Equities, Grim for Buck

Interesting week. Was that a sucker rally in the major indices or the start of a Christmas melt-up? Did this week mark the bottom for Treasuries yields? What's going on with commodities?

Did this cover (see right) in the wrong-way Economist contrarily mark the U.S. dollar bottom, as Stephen Vita suggested today on his Alchemy of Trading site? And why did a dozen wild turkeys just walk in front of my office window here in Quebec's Appalachians?

Lots of questions and plenty of interesting insights into it all, except the one about the turkeys, from this afternoon's Commitments of Traders report (based on data as of last Tuesday). I've just updated my "Latest Signals" page with the data from my trading setups based on these fascinating government reports, which are issued free each week by the U.S. Commodity Futures Trading Commission and detail trillions of dollars in futures and options holdings in nearly every market under the sun.

My overall take: the news is good for equity bulls. My COTs U.S. Composite Equity Index, based on my setups for the SP500, Dow Jones industrials, NASDAQ 100 and Russell 2000, remains unchanged from last week at 0.62. The index also gave a second renewed bullish signal this week. Renewed bullish signals for my Russell 2000 and Dow Jones industrials setups also lend more weight to the overall sunny picture here.

Also interesting: My 30-year Treasury setup hasn't followed suit with my 10-year Treasury setup, which you might recall turned bearish last week. In contrast, my 30-year bond setup has given its sixth straight renewed bullish signal. So I think it may still be an open question as to whether Treasury yields have really bottomed. This week's rebound may be but a pause. As I've noted this week, all my other Treasuries setups remain in bullish mode.

My odd-man-out 10-year setup is based on trading with the small traders when their net position hits historic extremes of bullishness or bearishness. The latest report shows this group of traders suddenly reducing their net position as a percentage of the total open interest. They've actually flipped back up to a decidedly bullish tilt in comparison with historic data, although not anywhere close enough to reverse this signal back to bullish, I should note.

As for the U.S. buck, the commercial futures traders have again slashed their net long position and now give me a renewed bearish signal. This is actually the first signal of any kind in this setup in over a year, since the original bearish signal back in Oct. 2006. It follows a steady 10-week reduction in the commercial net long position, which peaked with the Sept. 18 COTs report. How did Gisele Bundchen know!

I hope you had a good week and wish you a relaxing weekend. Check back here early next week to see more of my thoughts on the latest COTs data. Now, if only I could find my axe next time those turkeys show up.


1984 said...

A 3rd argument: The Other Dumb Money (TODM) is still SHORT SPX500 -- a Christmas rally would be as bad as soy. 500 trillion!? "Shit. I know shit's bad right now." ~President Camacho

Antony Pringle said...

COTS questions:

Thank you for an excellent blog and a very interesting discussion on the COTs. I was turned on to you blog from your posts on and I must admit that I’ve been somewhat obsessed with trying to duplicate your method over the past couple of weeks, with some success.

I do have some questions though:

· Re. Note 1: I noticed that you state that you system profits are based on buying upon an initial bullish (as opposed to renewed) signal, and shorting on an initial SELL. Am I right to assume that there are no intermediate cash positions (unless noted, like in sugar). So if you get a buy, like in the S&P500 on Sept 25, you buy and hold that position until you get a SELL signal, at which point you short? Have you run any scenarios for holding cash instead of shorting?
· Just wondering how you derived the Moving Average time and Standard Deviation number for each setup. Was it simply a question of crunching the numbers (brute force) for each combination of each until you found the most profitable combination? Any chance of getting the numbers you use for, say, you Agricultural Composite, CAD dollar, 13 wk Tbills and Crude Oil?
· I generated a setup for Nymex Gold and seem to be getting the same signals and COTs Index values as yours. However, I can’t figure out why you show a total of 17 trades for USERX, and 15 trades for HUI. Aren’t you using the same COTs data for both, and then just using the same signals to trade both? (Mine shows 17 trades, not including the first buy). Also, I noticed that using 2.1 standard deviations gives an even larger return, but in only 10 trades. Are you using 2.0 instead of 2.1 due to your validation test confidence interval?
· Could you point me somewhere to find information on calculating the T-test confidence intervals? I’m not sure that the T-test is appropriate since the population data (weekly net OI% values) are not normally distributed (lots of free normality check spreadsheets online). I’m not sure exactly what you’re comparing to arrive at the probabililites.

Again, thanks very much for a fascinating web site. As an Ottawa boy, living as an expat in Hong Kong, I also love your running commentary on the weather in the Eastern Townships! Enjoy your vacation. Any reply to this, large or small, is much appreciated.

Feel free to reproduce any part of this letter as a comment on your blog.


Tony Pringle

(more about me on my NEW blog at

Alex Roslin said...

Hi Tony,

Thanks for your message and interest in the subject. To answer you questions:

1) You're right. I'm working on some refined setups right now in which there would be cash positions.

2) The parameter values came from backtesting various combinations. I haven't published the values that aren't yet on the table, but may do so when I break off some of the setups for a paid-only area in the coming weeks.

3) My USERX setup uses different parameter values than the HUI and gold setup. Ten trades wouldn't be enough to qualify as statistically valid.

4) Wikipedia has some useful material on statistics, including the T-test. There are any number of statistical tests one can run, and this one seemed most appropriate, although I'm aware of the normality issue and am right now doing some other forms of testing to check the normalcy of the distributions, plus check on the robustness of the setups.

Good luck with your blog!